Figure 2.

The architecture of the proposed decision support system. Initially, individual models are trained using the datasets [clinical (CLIN), gene mutation (MUT), and gene expression (EXP)] with the three Machine Learning techniques [Random Forests (RF), Logistic Regression (LR), and Support Vector Machine (SVM)]. Subsequently, the best individual models are selected for each kind of data from all those produced (the asterisk symbol represents these models). In the following, a combination of these models is created by a classification committee with the final prediction vote weighted. Since the result of this committee and the models is a survival response, the process is repeated for all possible Treatment Intensity. In the end, the recommended therapy is the one that maximizes the probability of patient survival.